A Study Of Calibration Technology In Gaze Tracking System

With the rapid development of science and technology, the gaze tracking has been a hot direction in the field of scientific research, and it uses the eye feature which is got from eye images to estimate the line of sight. This technology has a wide range of application in human computer interaction, psychology, commercial application, safety and law enforcement and other fields.This paper first introduces the background of the research and significance of gaze tracking technology, and then introduces two existing gaze tracking technologies, which are the head-mounted based on two-dimensional and the desktop based on three-dimensional. Aiming at the problems existing in the key calibration technology in the eye tracking system, this paper studies two algorithms, and the experiments and simulation are designed to verify the availability of the algorithms.1. one calibration point method based on the two-dimensional head-mounted: as to the problem that the traditional nine calibration points method is too complex, this paper studies one calibration point method to simplify the calibration procedure; This method uses a local couple of points instead of a global to build the mapping relationship between the eye image and the scene image, improving the precision of the system. Experimental results show that: the precision of local one calibration point method is better than global and similar as the traditional nine calibration points method, but it simplifies the calibration procedure and allows the free head moving, therefore, the local one calibration point method is better than the traditional nine calibration points method.2. Binocular real time calibration algorithm based on the three-dimensional desktop: This paper studies a kind of real time calibration method based on the three-dimensional binocular model, through minimizing the distance between the intersections of the visual axes of the left and right eye with the surface of a display while subjects look naturally at the display to estimate the point of gaze in real time. This paper uses a simulation software to simulate the algorithm. The simulation results show that the algorithm allows head moving naturally in a certain range, moreover, using the algorithm to estimate one hundred points, validating that the algorithm is efficient. In the practical system, the mean estimation precision of the algorithm is 1.8 °. The precision of the method is low, and it needs to improve the precision further, but it reduces the complexity of calibration and improves the real-time of the system, so the real time calibration has great significance.